Query Optimization¶
Let the author state only what result a declarative query should return, then have a separate optimiser search the space of algebraically equivalent execution plans and emit the cheapest one under a cost model.
Core Idea¶
Query optimization is the database process that transforms a declarative query — what result is wanted, not how to compute it — into an execution plan chosen from a large space of algebraically equivalent candidates by a cost model. Its defining commitment separates specification from execution strategy: the author states logical intent in SQL, and a separate optimiser picks the physical procedure. Algebraic equivalence guarantees every candidate returns the identical relation, so the optimiser searches only for the cheapest.
Scope of Application¶
Lives wherever a declarative specification is rewritten over an algebraically-equivalent operator space against a cost model.
- Relational SQL optimizers — the home: System R, Volcano/Cascades, PostgreSQL, Oracle, SQL Server, MySQL.
- Stream-processing and big-data planners — Spark Catalyst, Flink, Trino/Presto over a different operator algebra.
- GraphQL and federated-API engines — a request graph rewritten into upstream calls and joins, costed by call cost and cardinality.
Clarity¶
Naming query optimization makes a division of labor legible: the author owns logical intent, the optimiser owns the physical procedure, and the DBA tunes the statistics and indexes the optimiser reads. A slow query then localizes cleanly — is the query wrong, the plan poor, or the statistics stale? — and foregrounding algebraic equivalence draws a firm line between reorderings that preserve the result and changes that alter it.
Manages Complexity¶
A join over a handful of tables admits 10^k candidate plans once join orders, operator implementations, and access paths are crossed. The optimizer collapses that space to one emitted plan by treating the candidates as a single equivalence class searched against one scalar, estimated cost. Algebraic equivalence factors correctness out entirely, so "find a fast, correct procedure" reduces to "find the cheapest member of an already-correct set."
Abstract Reasoning¶
The construct licenses a three-input diagnostic that localizes a fault to the query, the plan, or the statistics, and a change-the-optimizer's-mind interventionist move that adjusts cost-model inputs rather than the SQL. A correctness-preservation boundary rests on algebraic equivalence, and a feasibility move reasons about the super-exponential plan space as one equivalence class searched against a single cost scalar, predicting that plan quality is bounded by estimate accuracy.
Knowledge Transfer¶
Within data systems it transfers as mechanism, unchanged across relational optimizers, stream planners, and federated-API engines, because each is the same substrate — a declarative spec rewritten over an equivalent operator algebra against a cost model. Beyond that cluster the content decomposes into parents that genuinely recur: optimization, planning, and especially the emergent equivalence_preserving_rewriting (shared with compilers and proof rewriting). What stays home is the relational furniture — algebraic rules, catalog statistics, plan cache — so reaching "query optimization" toward route-finding is a stretched analogy carrying only optimization plus planning.
Relationships to Other Abstractions¶
Current abstraction Query Optimization Domain-specific
Parents (1) — more general patterns this builds on
-
Query Optimization is a kind of Optimization Prime
Query Optimization is optimization specialized to selecting the least-cost execution plan from behaviorally equivalent relational-algebra rewrites under a database cost model.
Hierarchy path (1) — routes to 1 parentless root
- Query Optimization → Optimization
Neighborhood in Abstraction Space¶
Query Optimization sits in a sparse region of the domain-specific corpus (75th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.
Family — Unclustered & Miscellaneous (309 abstractions)
Nearest neighbors
- Relational Model — 0.90
- Means-End Analysis — 0.83
- Graph Data Type — 0.82
- Mooers's law — 0.82
- Matrix — 0.81
Computed from structural-signature embeddings · 2026-07-12